SCinet: Testbed for High-Performance Networked Applications

نویسنده

  • William T. C. Kramer
چکیده

T he rapidly increasing availability of high-performance networking is arguably com-puting's most significant contribution to society in the past 20 years. Despite network availability, it is difficult to evaluate how well high-performance applications use its complex, multilevel interconnections in the real world. To better understand this interaction, leading experts and organizations in ultra-high-performance networking and computing come together once a year to build what is likely the world's most intense, diverse, high-performance network. Almost overnight, the SCinet testbed comes online to showcase state-of-the-art network technology at the IEEE/ACM-sponsored supercomputing conference known as SCxy. The testbed is in place for four days, during which time teams of application developers demonstrate bandwidth-intensive applications that stress the network's capabilities. At the end of the demonstrations, SCinet is dismantled, and work begins on the next year's design. SCinet runs hundreds of bandwidth-intensive applications each year, but SC2000 and SC2001 featured a special " Network Bandwidth Challenge " for applications to try making full use of SCinet's wide-area capabilities. In the words of one network engineer, the Bandwidth Challenges asked application developers to " burn down the world's fastest network. " The SCinet team selected 10 of the most data-intensive applications in 2000 and 12 applications in 2001 for formal evaluation, and Qwest provided prizes for the winners. SCinet is built from loaned networking equipment and services—more than $25 million each year for 2000 and 2001, as well as the donated effort of more than 100 leading network engineers. The SC2000 network was created and fully operational in just over five days, and consisted of 82 miles of fiber optic cable installed in less than 51 hours. SC2001 took similar amounts of fiber and time. At the SC2000 conference in Dallas, the local area network (LAN) within the SCinet testbed area had a peak capacity of 130 gigabits per second. The network connected to all the major national scientific networks and supercomputer centers, and maintained a total wide-area network (WAN) band-width of almost 9 Gbps. At SC2001 in Denver, the LAN capacity increased by 28 percent—to 194 Gbps—and the WAN capacity almost doubled—to 15.7 Gbps. SCinet is an Internet service provider in its own right. Its network design is intentionally complex to explore issues typically encountered in real-world networking, such as interoperating between different network domains and across different routers and technologies. The overall network design for SC2000 and SC2001 consisted of four network levels …

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Designinga Neuro-Sliding Mode Controller for Networked Control Systems with Packet Dropout

This paper addresses control design in networked control system by considering stochastic packet dropouts in the forward path of the control loop. The packet dropouts are modelled by mutually independent stochastic variables satisfying Bernoulli binary distribution. A sliding mode controller is utilized to overcome the adverse influences of stochastic packet dropouts in networked control system...

متن کامل

Designing Advanced Research Computing Academic Programs

We present an overview of current academic curricula for Scientific Computing, High-Performance Computing and Data Science. After a survey of current academic and non-academic programs across the globe, we focus on Canadian programs and specifically on the education program of the SciNet HPC Consortium, using its detailed enrollment and course statistics for the past four to five years. Not onl...

متن کامل

Resource Allocation Algorithms for Event - Based Enterprise Systems by Alex King Yeung Cheung

Resource Allocation Algorithms for Event-Based Enterprise Systems Alex King Yeung Cheung Doctor of Philosophy Graduate Department of Electrical and Computer Engineering University of Toronto 2011 Distributed event processing systems suffer from poor scalability and inefficient resource usage caused by load distributions typical in real-world applications. The results of these shortcomings are a...

متن کامل

A networked mobile sensor test-bed for collaborative multi-target tracking applications

This paper presents the design, architecture, implementation, and experimental results from a networked mobile sensor test-bed developed for collaborative sensor tracking applications. The test-bed comprises a fleet of networked mobile sensors, an indoor localization system, a control, debugging and management infrastructure, and a tiered wireless ad hoc network for seamless integration of the ...

متن کامل

Emulation Testbed for IEEE 802.15.4 Networked Systems

IEEE 802.15.4 based devices are a key component for mobile and pervasive computing. However, their small dimensions and reduced resources, together with the intrinsic properties of wireless communication, make it difficult to evaluate such networked systems through realworld trials. In this paper we present an emulation testbed intended for the evaluation of IEEE 802.15.4 networked systems. The...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

عنوان ژورنال:
  • IEEE Computer

دوره 35  شماره 

صفحات  -

تاریخ انتشار 2002